Using RTT Variability for Adaptive Cross-Layer Approach to Multimedia Delivery in Heterogeneous Networks

• Published in: IEEE transactions on multimedia 2009-10, Vol.11 (6), p.1194-1203
• Main Authors: Choi, Baek-Young, Song, Sejun, Wang, Yue, Park, Eun Kyo
• Format: Article
• Language: English
• Subjects: Access protocols; Automatic control; Channels; Cities and towns; Constraining; Cross layer design; Cross-layer approach; Forward error correction; heterogenous networks; Multimedia; Networks; Optimization; Performance analysis; Product introduction; Propagation losses; Proposals; Quantitative analysis; round-trip time (RTT) variability; Standard deviation; Streaming media; Wireless communication; Wireless networks
• ISSN: 1520-9210; 1941-0077
• DOI: 10.1109/TMM.2009.2026103

A holistic approach should be made for a wider adoption of a cross-layer approach. A cross-layer design on a wireless network assumed with a certain network condition, for instance, can have a limited usage in heterogeneous environments with diverse access network technologies and time varying network performance. The first step toward a cross-layer approach is an automatic detection of the underlying access network type, so that appropriate schemes can be applied without manual configurations. To address the issue, we investigate the characteristics of round-trip time (RTT) on wireless and wired networks. We conduct extensive experiments from diverse network environments and perform quantitative analyses on RTT variability. We show that RTT variability on a wireless network exhibits greatly larger mean, standard deviation, and min-to-high percentiles at least 10 ms bigger than those of wired networks due to the MAC layer retransmissions. We also find that the impact of packet size on wireless channel is particularly significant. Thus through a simple set of testing, one can accurately classify whether or not there has been a wireless network involved. We then propose effective adaptive cross-layer schemes for multimedia delivery over error-prone links. They include limiting the MAC layer retransmissions, controlling the application layer forward error correction (FEC) level, and selecting an optimal packet size. We conduct an analysis on the interplay of those adaptive parameters given a network condition. It enables us to find optimal cross-layer adaptive parameters when they are used concurrently.